Forced Subunit Assembly in α1β2γ2GABAAReceptors

The major isoform of the γ-aminobutyric acid type A (GABAA) receptor is thought to be composed of 2α1, 2β2, and 1γ2 subunit(s), which surround the ion pore. Definite evidence for the subunit arrangement is lacking. We show here that GABAA receptor subunits can be concatenated to a trimer that can be functionally expressed upon combination with a dimer. Many combinations did not result in the functional expression. In contrast, four different combinations of triple subunits with dual subunit constructs, all resulting in the identical pentameric receptor γ2β2α1β2α1, could be successfully expressed in Xenopus oocytes. We characterized the functional properties of these receptors in respect to agonist, competitive antagonist, and diazepam sensitivity. All properties were similar to those of wild type α1β2γ2 GABAAreceptors. Thus, together with information on the crystal structure of the homologous acetylcholine-binding protein (Brejc, K., van Dijk, W. J., Klaassen, R. V., Schuurmans, M., van Der Oost, J., Smit, A. B., and Sixma, T. K., (2001) Nature411, 269–276, we provide evidence for an arrangement γ2β2α1β2α1, counterclockwise when viewed from the synaptic cleft. Forced subunit assembly will also allow receptors containing different subunit isoforms or mutant subunits to be expressed, each in a desired position. The methods established here should be applicable to the entire ion channel family comprising nicotinic acetylcholine, glycine, and 5HT3 receptors.

The inferred arrangement of subunits around the channel pore is hypothetical, based on the findings that the GABA-binding site is located at intersubunit contacts between ␣ and ␤ subunits (17)(18)(19)(20)(21) and that homologous amino acid residues of ␣ and ␥ subunits form the benzodiazepine-binding pocket (22)(23)(24)(25)(26)(27)(28). The observation that assembly intermediates comprising ␣␥ or ␣␤ dimers displayed some benzodiazepine or agonist binding, respectively (29), supported conclusions drawn from the former mutation studies. From the crystal structure of the acetylcholine-binding protein (30), a protein homologous to the extracellular domain of the nicotinic acetylcholine receptors and the other members of the superfamily of ligand-gated ion channels, we can deduce the absolute position of the amino acid residues involved in the formation of agonist and drug-binding sites in a subunit in the pentamer. However, this acetylcholinebinding protein is a homopentamer and gives no information about the arrangement of the heteromeric GABA A receptors.
In a former study (31) we expressed ␣-␤ and ␤-␣ tandem constructs in combination with single ␥ subunits in Xenopus oocytes and investigated the function of the formed receptors. The results suggested a possible arrangement ␥␤␣␤␣; however, some uncertainty remained. From this work it also was not clear whether exclusively one or several arrangements of a given set of subunits is possible.
In the present study we aimed to express GABA A receptors from various combinations of linked constructs containing two and three subunits. A single pentameric arrangement made of ␣ 1 , ␤ 2 , and ␥ 2 subunits was found to result in functional ion channels. We describe its properties toward the agonist GABA, the competitive antagonist bicuculline, and the positive allosteric modulator diazepam. The described techniques will allow forced assembly and functional study of GABA A and related receptors of defined subunit arrangements.
Two-Electrode Voltage-Clamp Measurements-All measurements were done in medium containing 90 mM NaCl, 1 mM MgCl 2 , 1 mM KCl, 1 mM CaCl 2 , and 5 mM HEPES, pH 7.4, at a holding potential of Ϫ80 mV. For the determination of maximal current amplitudes 1 mM GABA (Fluka, Buchs, Switzerland) was applied for 20 s. Voltage-dependent sodium currents were determined by a potential jump from a holding potential of Ϫ100 to Ϫ15 mV. As the modes of activation of the GABA receptor channel and the voltage-dependent sodium channel differ, the measurements of the two channels do not interfere with each other. The GABA-evoked peak current amplitude was standardized to the coexpressed sodium current amplitude of the same oocyte. The mean standardized current amplitude of at least 3 oocytes per subunit combination was then compared with the mean standardized wild type current amplitude. GABA-evoked currents (at 8 -12% of the maximal current amplitude) were inhibited by varying concentrations of bicuculline methiodide (RBI). Relative current stimulation by diazepam was determined at a GABA concentration evoking 2-5% of the maximal current amplitude in combination with varying concentrations of diazepam (DZ) (Roche) and expressed as ((I (GABAϩDZ) /I (GABA) )Ϫ1)ϫ100%.

Engineering of Functional Triple Subunit Constructs-We
have shown previously (31) that it is feasible to covalently link ␣ and ␤ subunits of the GABA A receptor while retaining full receptor function (31). These linked constructs allowed us to propose a possible arrangement of subunits in the ␣ 1 ␤ 2 ␥ 2 receptor. However, some uncertainty remained due to a possible rearrangement of dual subunit constructs. We now have linked three subunits in different sequence and expressed them in combination with tandem constructs. Triple subunit constructs unlike dual subunit constructs cannot rearrange for topological reasons.
In the following, the sequence of subunits in multiple subunit constructs is always described as the C-terminal of the first subunit linked to the N-terminal of the second subunit. To link the ␥ subunit to either side of the existing tandem constructs ␣-10-␤ and ␤-23-␣, linker lengths for the new connections had to be established. To test functionality of these linkers, dual constructs ␣-␥, ␥-␣, ␤-␥, or ␥-␤ were co-expressed with ␣-␤ or ␤-␣ constructs and single ␣ or ␤ subunits to yield receptors of the composition 2␣ 2 ␤ 1 ␥, e.g. ␥-␤/␣-␤/␣ (data not shown). The required linker lengths for functional expression were found to be 26 amino acid residues for ␥-␤, 10 amino acid residues for ␣-␥, and 23 residues for ␥-␣. For a ␤-␥ tandem construct, linkers up to 25 amino acid residues in length were tested but were found to result only in very small functional channel expression. In the following, the triple constructs Expression of Receptors with Different Arrangements of Linked Subunits-If the presence of 2␣, 2␤, and 1␥ subunit(s) in a pentamer is assumed, six different arrangements of ␣, ␤, and ␥ subunits are possible (Fig. 1). Of this stoichiometry we tested first arrangements containing one ␥ and two each of ␣ and ␤ in an alternating fashion. Assembly studies of Tretter et al. (15) suggest such an alternating arrangement because expression of either ␣ and ␥ or ␤ and ␥ leads to dimers only. In these cases either a ␤ or ␣ subunit is missing, respectively, to continue assembly. Expression of ␣ and ␤, in contrast, leads to the formation of tetra-and pentamers. Two alternating arrangements are possible, namely ␥␤␣␤␣ and ␥␣␤␣␤. They are the most probable arrangements as they seem to contain the inferred two ␣␤ and ␣␥ subunit interfaces required for the formation of two GABA-and one benzodiazepine-binding sites. As the subunits are not symmetrical, only one of these two alternating arrangements is predicted to form the correct subunit interfaces for establishing binding sites.
The fact that no or very small currents were observed during linker length optimization and for many of the above mentioned subunit combinations indicates that proteolysis in the linker regions is not occurring to a significant extent. Small currents as observed, for example, for ␥-␤-␣/␣ and ␥-␤-␣/␤ can be thought to reflect mis-assembled channels. These channels are not necessarily silent, but are not formed to a significant extent.
Concentration-response Properties of Receptors Made from Linked Constructs-To characterize receptors with the subunit arrangement ␥␤␣␤␣ made from linked constructs we investigated their response properties to the agonist GABA and the competitive antagonist bicuculline. First, we studied the GABA concentration-response properties of the tandem construct ␤-␣ in combination with single ␥ 2 subunits. Compared with receptors made from single subunits (wild type) we observed a slight rightward shift (2.4-fold; Fig. 5B). A similar 2-fold shift had already been observed for the combination of ␤-␣ with single ␤ subunits (31). Next, three of four combinations of triple constructs with tandem constructs that were proven functional before (Fig. 3) were characterized in their GABA concentrationresponse properties. Fig. 5A shows representative current traces obtained with the subunit combination. All of them behaved very similarly, showing properties similar to wild type receptors regarding EC 50 values ( Fig. 5 and Table I). ␤-␣-␤/␣-␥ Forced Subunit Assembly in ␣ 1 ␤ 2 ␥ 2 GABA A Receptors was only analyzed twice and had an average EC 50 of 124 M in these experiments. Fig. 6 shows a summary of inhibition experiments by the competitive inhibitor bicuculline of GABA-induced currents on all subunit construct combinations resulting in functional ␥␤␣␤␣ receptors except ␤-␣-␤/␣-␥. The obtained bicuculline concentration-response curves were very similar to those of the receptor made from single subunits. Table I summarizes the agonist and competitive antagonist properties of the different constructs. From these results we can conclude that the linkers used here between the subunits have little influence on the apparent affinity for GABA induced channel opening and the inhibition by bicuculline.
Diazepam Responsiveness of Receptors Made from Linked Constructs-It has earlier been observed that stimulation by diazepam of currents elicited by GABA in oocytes expressing ␣:␤:␥ ϭ 1:1:1 is smaller and more variable than in oocytes expressing ␣:␤:␥ ϭ 1:1:5 (36). Therefore, it was interesting to examine receptors formed from triple and tandem constructs in this respect. The response to increasing concentrations of diazepam did not markedly differ from receptors containing loose subunits (Fig. 7) regarding the concentration-dependence of current stimulation. However, there was a difference regarding maximal stimulation and variability in different oocytes of this value. Although stimulation by diazepam of currents elicited by GABA in oocytes expressing ␣:␤:␥ ϭ 1:1:5 centered at about 170% and was quite variable in different oocytes, a value of about 270% with little variation was observed, provided the ␥ 2 subunit was covalently linked to other subunits to give ␥-␤-␣/ ␤-␣. Fig. 8 documents this by comparing diazepam stimulation in oocytes either expressing ␣:␤:␥ ϭ 1:1:5 or covalently linked constructs ␥-␤-␣:␤-␣ ϭ 1:1. The combinations ␤-␣-␥/␤-␣, ␣-␤-␣/ ␥-␤, and ␤-␣/␥ also showed values for maximal stimulation, which were higher than those found for wild type receptors and had small variations (data not shown). It has been observed earlier that the stimulation by diazepam decreases in oocytes injected with cRNA coding for ␣, ␤, and ␥ during expression time (36). This phenomenon seemed not to occur in oocytes expressing linked subunits (not shown).

DISCUSSION
In the present study we used covalently linked subunits of the GABA A receptor to study the arrangement of subunits in ␣ 1 ␤ 2 ␥ 2 receptors. We examined receptors made from combinations of triple and dual subunit constructs containing ␣ 1 , ␤ 2 , and/or ␥ 2 subunits in different orders. We present direct evidence that ␣ 1 ␤ 2 ␥ 2 receptors have a subunit composition and relative arrangement of ␥␤␣␤␣ from N-terminal to C-terminal. It should be stated that we restricted ourselves to looking at the ability of the subunit combinations to form functional ion channels. For combinations that did not result in function, we cannot say at present whether this is due to insufficient stability of constructs, or due to assembly problems, or whether the receptors reach the surface membrane and are unable to open.
The majority of GABA A receptors contains ␣, ␤, and ␥ subunits (9,(37)(38)(39)(40). This subunit composition can be reached by combining either three subunits of one type with one subunit each of the other two types or two subunits each from two types with one subunit of the third type. The former stoichiometry has been excluded for ␣ 3 ␤ 2 ␥ 2 and ␣ 1 ␤ 2 ␥ 2 receptors by electrophysiological characterization of mutant receptors (13,14). For FIG. 5. A, representative current traces of a GABA concentrationresponse experiment for the construct combination ␥-␤-␣/␤-␣. Digitized traces were recorded using MacLab (ADInstruments). The vertical bars above the traces represent the duration of the GABA application. B, GABA concentration-response curves of ␣ 1 ␤ 2 ␥ 2 receptors made from single subunits (f), ␤-␣ tandem constructs co-expressed with single ␥ 2 subunits (q), triple construct ␥-␤-␣ combined with ␤-␣ (OE), ␤-␣-␥ combined with ␤-␣ (), and ␣-␤-␣ combined with ␥-␤ (ࡗ). All curves obtained from construct combinations are very similar compared with those from receptors made from single subunits. Mean values with S.E. from 4 -5 oocytes from two batches for each subunit combination are shown. Individual curves were first normalized to the observed maximal current amplitude and subsequently averaged.  (14), investigations using immunoprecipitation (15), and fluorescence energy transfer studies (16). However, it has been reported that also two ␥ subunits can occur within the same pentamer. This possibility has been suggested on the basis of co-occurrence of isoforms of the ␥ subunit within the same receptor molecule, where ␥ 2 ␥ 3 and ␥ 2S ␥ 2L pairings have been detected in immunoprecipitation experiments (41,42). We found that only combinations of triple and dual constructs, which resulted in the arrangement ␥␤␣␤␣, therefore containing 2␣, 2␤, and 1␥ subunit(s), were functionally expressed with properties very similar to receptors made from single subunits. Combinations of subunits and/or linked subunit constructs that would result in a different stoichiometry or a different arrangement were not functional. Our study is limited to ␣ 1 , ␤ 2 , and ␥ 2 containing receptors. It is not clear whether receptors containing different isoforms of these subunit types have the same stoichiometry and arrangement as the ␣ 1 ␤ 2 ␥ 2 receptor. Taking into account that for the formation of GABA-binding sites defined interfaces must be formed, it is likely that all ␣␤␥, ␣␤␦, and ␣␤⑀ (4, 43) receptors follow the same building plan. When we expressed triple constructs in combination with tandem constructs to form ␥␤␣␤␣ receptors we observed concentration-response properties of the resulting receptors very similar to receptors made from single, individual subunits. The linkers between the subunits do not strongly affect function. Only in one case did we observe a slight decrease in the Hill coefficient. This coefficient is 1.0 in the case of ␥-␤-␣/␤-␣ as compared with 1.2-1.4 for all other receptors (Table I) 2␤) turn in the same direction, whereas 1␤ subunit turns in the opposite direction. This ␤ subunit can be thought to be replaced by the ␥ subunit in the ␣␤␥ receptor. In this case the linker between ␥ and ␤ in the ␥-␤-␣ construct could impair the suggested movement of the linked subunits, which could lead to the observed decrease of the Hill coefficient. This movement, however, did not seem to be disturbed in the ␥-␤ dual construct when expressed in combination with the ␣-␤-␣ triple construct (not shown). A dual subunit construct might be more flexible than a triple construct. Introduction of a shorter linker into the triple construct could confirm the above hypothesis.
For all receptors formed from linked subunits we observed a 2.0 -3.5-fold shift to the right in the GABA concentration-response curves compared with receptors made from single subunits. On one hand this shift could be due to changed binding or gating properties of the receptor pentamers by linkage. On the other hand it has been observed frequently that expression of single ␣, ␤, and ␥ subunits in oocytes or HEK cells leads to a mixed population of ␣␤␥ and ␣␤ receptors (36). The EC 50 of ␣␤ receptors is about 8 M, that of ␣␤␥ receptors about 41 M (45). Expression of linked constructs does not allow the formation of pentameric ␣␤ receptors, and the 2.0 -3.5-fold shift of the concentration-response curve further to the right could at least be partly due to expression of pure ␣␤␥ receptors. This view is supported by the markedly higher stimulation of GABA-induced currents by diazepam of receptors made from linked subunits. Presence of ␣␤ in ␣␤␥ receptors decreases apparent diazepam stimulation (36).
Our approach leads to the relative sequence of subunits in the receptor only, and does not allow a statement about the absolute arrangement, e.g. the sequence when viewed from the synaptic cleft. The determination of the crystal structure of the acetylcholine-binding protein (30) leads to an insight into the absolute configuration of the extracellular parts of the subunits and interfaces of the nicotinic acetylcholine receptor. The acetylcholine-binding protein is a bacterial protein homologous to the extracellular domain of the nicotinic acetylcholine receptor of higher organisms. The GABA A receptor belongs to the same superfamily of ligand-gated ion channels and is therefore structurally homologous to the nicotinic acetylcholine receptor. The knowledge of amino acid residues that form agonist and benzodiazepine-binding sites, respectively (for review see Ref. 27) allows us to conclude on which side of the ␣ subunit ␥ and ␤ subunits are positioned when viewed from the synaptic cleft. Therefore, the absolute arrangement of the ␣ 1 ␤ 2 ␥ 2 receptor is very likely ␥␤␣␤␣ from the N terminus to the C terminus, read in anti-clockwise direction when viewed from the synaptic cleft (Fig. 9).
This work will allow the study of the roles of any individual site located on an ␣ or ␤ subunit, which was impossible thus far because there were always two sites on the two subunits affected by a mutation in 2␣2␤1␥ GABA A receptors. The possibility of a forced subunit assembly will enable targeted introduction of a mutation in only one subunit. This will, for example, allow dissection of the two low affinity agonist sites located at the ␣␤ subunit interface. Forced subunit assembly will have further impact on the characterization of receptor forms containing different isoforms of the same subunit subtypes. Receptors that are made of 4 or 5 different subunits cannot be analyzed by recombinant expression of the mixture of the single subunits because many different receptor subtypes can be formed. Expression of predefined sequences of subunits in triple and tandem constructs will allow the study of pharmacological properties of defined receptor isoforms. As the GABA A receptor belongs to a superfamily of ligand-gated ion channels, methods described here are applicable to neuronal and non-neuronal nicotinic acetylcholine receptors, glycine receptors, and 5HT 3 receptors.